CN108369534B

CN108369534B - Code execution request routing

Info

Publication number: CN108369534B
Application number: CN201680072794.XA
Authority: CN
Inventors: 斯科特·丹尼尔·魏斯涅夫斯基; 马克·约翰·布鲁克; 蒂莫西·艾伦·瓦格纳
Original assignee: Amazon Technologies Inc
Current assignee: Amazon Technologies Inc
Priority date: 2015-12-21
Filing date: 2016-12-15
Publication date: 2022-03-11
Anticipated expiration: 2036-12-15
Also published as: JP6647398B2; EP3394750B1; US20180210760A1; US20190227849A1; CN108369534A; WO2017112526A1; US11016815B2; JP2018537776A; JP6892729B2; US20170177413A1; JP2020074153A; US9910713B2; EP3394750A1; US10248467B2

Abstract

A system for providing low-latency computing capacity is provided. The system may be configured to route incoming code execution requests based on user direction using a particular container running on multiple virtual machine instances. The system may be configured to process a code execution request, identify based on a user indication that a particular container previously used to process similar types of requests is to be used to process the code execution request, and cause the code execution request to be processed using the particular container.

Description

Code execution request routing

Background

Generally, computing devices exchange data using a communication network or a series of communication networks. Companies and organizations operate computer networks that interconnect multiple computing devices to support operations or provide services to third parties. Computing systems may be located in a single geographic location or in multiple different geographic locations (e.g., interconnected via a private or public communication network). In particular, a data center or data processing center, collectively referred to herein as a "data center," may include a plurality of interconnected computing systems to provide computing resources to users of the data center. The data center may be a private data center operated on behalf of an organization, or may be a public data center operated on behalf of or for the benefit of the public.

To facilitate increased utilization of data center resources, virtualization techniques may allow a single physical computing device to host one or more instances of a virtual machine that appear to users of the data center and operate as a stand-alone computing device. With virtualization, a single physical computing device may create, maintain, delete, or otherwise manage virtual machines in a dynamic manner. In turn, users may request computer resources from the data center, including configurations of single computing devices or networked computing devices, and may be provided with different amounts of virtual machine resources.

In some scenarios, a virtual machine instance may be configured according to multiple virtual machine instance types to provide specific functionality. For example, various computing devices may be associated with different combinations of operating systems or operating system configurations, virtualized hardware resources, and software applications to enable the computing device to provide different desired functionality, or to more efficiently provide similar functionality. These virtual machine instance type configurations are typically contained within a device image that includes static data that contains the software (e.g., OS and applications, their configuration and data files, etc.) that the virtual machine runs upon startup. The device image is typically stored on a disk for creating or initializing the instance. Thus, the computing device may process the device image to implement the desired software configuration.

Drawings

The foregoing aspects and many of the attendant advantages of this disclosure will become more readily appreciated as the same become better understood by reference to the following detailed description, when taken in conjunction with the accompanying drawings, wherein:

FIG. 1 is a block diagram depicting an illustrative environment for providing low-latency computing capacity, in accordance with exemplary aspects;

FIG. 2 depicts a general architecture of a computing device providing a routing manager for routing code execution requests, according to an exemplary aspect;

FIG. 3 is a flow diagram illustrating a code execution request routing routine implemented by a routing manager in accordance with an exemplary aspect;

FIG. 4 depicts an exemplary mapping table showing routing of requests to containers in accordance with an exemplary aspect;

FIG. 5 is a flow diagram illustrating a code execution result caching routine implemented by a routing manager in accordance with an illustrative aspect; and

FIG. 6 is a flow diagram illustrating a container lifecycle management routine implemented by a routing manager, according to an exemplary aspect.

Detailed Description

Companies and organizations no longer need to acquire and manage their own data centers in order to perform computing operations (e.g., execute code, including threads, programs, functions, software, routines, subroutines, procedures, etc.). With the advent of cloud computing, the storage space and computing power traditionally provided by hardware computing devices is now available and configurable over the internet in a few minutes. Thus, a developer can quickly purchase a desired amount of computing resources without having to worry about acquiring a physical machine. These computing resources are typically purchased in the form of virtual computing resources or virtual machine instances. These examples of virtual machines are software implementations of physical machines (e.g., computers) that are hosted on physical computing devices and may contain operating systems and applications traditionally provided on physical machines. These virtual machine instances are configured using a set of computing resources (e.g., memory, CPU, disk, network, etc.) that an application running on the virtual machine instance can request and utilize in the same manner as a physical computer.

However, even when purchasing virtual computing resources, developers still have to decide how many and what types of virtual machine instances to purchase, and how long to keep them. For example, the cost of using virtual machine instances may vary depending on their lease type and number of hours. In addition, the minimum time a virtual machine can rent is typically on the order of a few hours. In addition, the developer must specify the hardware and software resources (e.g., type of operating system and language runtime, etc.) to be installed on the virtual machine. Other problems they may encounter include over-utilization (e.g., acquiring too little computing resources and suffering performance problems), under-utilization (e.g., acquiring more computing resources than needed to run code and therefore paying too much), traffic variation prediction (e.g., so that they know when to scale up or down), and instance and language runtime startup delays, which may take 3-10 minutes or more, even though a user may desire computing capacity on the order of seconds or even milliseconds.

According to aspects of the present disclosure, by maintaining a pool of pre-initialized virtual machine instances that are ready for use upon receipt of a user request, and automatically managing the amount of capacity available in the pool to service incoming requests, the delay (sometimes referred to as latency) associated with executing user code (e.g., instance and language runtime start-up time) may be significantly reduced, and utilization may be improved.

In general, systems and methods are disclosed that facilitate managing virtual machine instances in a virtual computing system. A virtual computing system maintains a pool of virtual machine instances on which one or more software components (e.g., operating systems, language runtimes, libraries, etc.) are loaded. Maintaining a pool of virtual machine instances can involve creating new instances, obtaining new instances from an external instance provisioning service, destroying instances, assigning/reassigning instances to users, modifying instances (e.g., containers or resources therein), and so forth. The virtual machine instances in the pool may be designated to service user requests to execute program code. In this disclosure, the phrases "program code," "user code," and "cloud function" may sometimes be used interchangeably. The program code may execute in a separate container created on the virtual machine instance. Because the virtual machine instances in the pool have already been launched and loaded with a particular operating system and language runtime when the request is received, the latency associated with finding the computing capacity (e.g., by executing user code in one or more containers created on the virtual machine instances) that can handle the request is significantly reduced.

In another aspect, a virtual computing system may create and manage a mapping between incoming code execution requests received by the virtual computing system and computing capacity for processing those code execution requests. The mapping may facilitate multiple requests to use and reuse certain container-specific resources. For example, a given container may be associated with a cache that stores results of code execution and may be accessed by any program code executing in the given container. Thus, once the code execution results associated with a request are stored in the cache, subsequent requests of the same type may be more efficiently processed using the code execution results stored in the cache if they are routed to a given container. Thus, by routing the request to the appropriate container running on the virtual computing system, latency gains can be achieved.

Specific embodiments and exemplary applications of the present disclosure will now be described with reference to the accompanying drawings. These embodiments and exemplary applications are intended to illustrate, but not to limit the disclosure.

Illustrative Environment including virtual computing System

Referring to fig. 1, a block diagram illustrating an embodiment of a virtual environment 100 will be described. The example shown in fig. 1 includes a virtualized environment 100 in which a user (e.g., a developer, etc.) of a user computing device 102 may run various program code using virtual computing resources provided by a virtual computing system 110.

By way of illustration, various exemplary user computing devices 102 (including desktop computers, laptop computers, and mobile phones) are shown in communication with a virtual computing system 110. In general, the user computing device 102 may be any computing device, such as a desktop computer, a laptop computer, a mobile phone (or smart phone), a tablet, a kiosk, a wireless device, and other electronic devices. Additionally, the user computing device 102 may include web services running on the same or different data centers, where, for example, different web services may programmatically communicate with each other to perform one or more of the techniques described herein. Further, the user computing device 102 may include internet of things (IoT) devices, such as internet appliances and connected devices. The virtual computing system 110 may provide the user computing device 102 with one or more user interfaces, Command Line Interfaces (CLIs), Application Programming Interfaces (APIs), and/or other programming interfaces, all for generating and uploading user code, invoking user code (e.g., submitting requests to execute user code on the virtual computing system 110), scheduling event-based or timed jobs, tracking user code, and/or viewing other log or monitoring information related to its requests and/or user code. Although one or more embodiments may be described herein as using a user interface, it should be understood that these embodiments may additionally or alternatively use any CLI, API, or other programming interface.

The user computing device 102 accesses the virtual computing system 110 over the network 104. The network 104 may be any wired network, wireless network, or combination thereof. Additionally, the network 104 may be a personal area network, a local area network, a wide area network, an over-the-air broadcast network (e.g., for radio or television), a cable network, a satellite network, a cellular telephone network, or a combination thereof. For example, the network 104 may be a publicly accessible network linking networks, possibly operated by various parties (such as the internet). In some embodiments, the network 104 may be a private or semi-private network, such as a corporate or university intranet. The network 104 may include one or more wireless networks, such as a global system for mobile communications (GSM) network, a Code Division Multiple Access (CDMA) network, a Long Term Evolution (LTE) network, or any other type of wireless network. The network 104 may use protocols and components to communicate via the internet or any other of the types of networks described above. For example, the protocols used by the network 104 may include hypertext transfer protocol (HTTP), HTTP Secure (HTTPs), Message Queue Telemetry Transport (MQTT), constrained application protocol (CoAP), and the like. Protocols and components for communicating via the internet or any other above-mentioned type of communication network are well known to those skilled in the art and are therefore not described in detail herein.

Virtual computing system 110 is depicted in FIG. 1 as operating in a distributed computing environment comprising several computer systems interconnected using one or more computer networks. The virtual computing system 110 may also operate within a computing environment having a fewer or greater number of devices than shown in FIG. 1. Accordingly, the description of virtual computing system 110 in FIG. 1 should be taken as illustrative and not limiting of the present disclosure. For example, virtual computing system 110, or various components thereof, may implement various web service components, hosted or "cloud" computing environments, and/or peer-to-peer network configurations to implement at least a portion of the processes described herein.

Furthermore, virtual computing system 110 may be implemented in hardware and/or software, and may, for example, include one or more physical or virtual servers implemented on physical computer hardware configured to execute computer-executable instructions for performing various features that will be described herein. One or more servers may be geographically dispersed or geographically co-located, for example, in one or more data centers.

In the environment shown in FIG. 1, the virtual environment 100 includes a virtual computing system 110 that includes a front end 120, a warm-up pool manager 130, a worker manager 140, and a routing manager 150. In the depicted example, virtual machine instances ("instances") 152, 154 are shown in a warm-up pool 130A managed by warm-up pool manager 130, and instances 156, 157, 158, 159 are shown in an active pool 140A managed by worker manager 140. In some embodiments, the term "virtual machine instance" may refer to execution of software or other executable code that emulates hardware to provide an environment or platform ("execution environment") on which software may be executed. A virtual machine instance is typically executed by a physical hardware device, which may be different from the hardware emulated by the virtual machine instance. For example, the virtual machine may emulate a first type of processor and memory when executing on a second type of processor and memory. Thus, a virtual machine may be utilized to execute software intended for a first execution environment (e.g., a first operating system) on a physical device executing a second execution environment (e.g., a second operating system). In some cases, the hardware emulated by the virtual machine instance may be the same as or similar to the hardware of the underlying device. For example, an apparatus having a first type of processor may implement multiple virtual machine instances, each virtual machine instance emulating an instance of the first type of processor. Thus, a single device may be divided into multiple logical sub-devices (each referred to as a "virtual machine instance") using a virtual machine instance. While virtual machine instances can typically provide a level of abstraction that is remote from the underlying physical device hardware, such abstraction is not required. For example, assume that a device implements multiple virtual machine instances, where each virtual machine instance emulates the same hardware as that provided by the device. In this scenario, each virtual machine instance may allow software applications to execute code on the underlying hardware without translation while maintaining logical separation between software applications running on other virtual machine instances. This process, commonly referred to as "native execution," may be used to increase the speed or performance of a virtual machine instance. Other techniques are known in the art that allow direct utilization of the underlying hardware, such as hardware pass-through techniques.

The illustration of the various components within the virtual computing system 110 is logical in nature, and one or more components may be implemented by a single computing device or multiple computing devices. For example, instances 152, 154, 156, 157, 158, 159 may be implemented on one or more physical computing devices located in various different geographic regions. Similarly, each of the front end 120, the warming pool manager 130, the worker manager 140, and the routing manager 150 may be implemented on multiple physical computing devices. Alternatively, one or more of the front end 120, the warming pool manager 130, the worker manager 140, and the routing manager 150 may be implemented on a single physical computing device. In some embodiments, the virtual computing system 110 may include multiple front ends, multiple warm-up pool managers, multiple worker managers, and/or multiple capacity managers. Although six virtual machine instances are shown in the example of fig. 1, the embodiments described herein are not so limited, and those skilled in the art will appreciate that virtual computing system 110 may include any number of virtual machine instances implemented using any number of physical computing devices. Similarly, although a single warm-up pool and a single active pool are shown in the example of fig. 1, the embodiments described herein are not so limited, and one skilled in the art will appreciate that the virtual computing system 110 may include any number of warm-up and active pools.

In the example of fig. 1, virtual computing system 110 is shown connected to network 104. In some embodiments, any component within the virtual computing system 110 may communicate with other components of the virtual environment 100 (e.g., the user computing device 102 and the auxiliary service 106, which may include a monitoring/logging/billing service 107, a storage service 108, an instance provisioning service 109, and/or other services that may communicate with the virtual computing system 110) via the network 104. In other embodiments, not all components of the virtualized computing system 110 are able to communicate with other components of the virtualized environment 100. In one example, only the front end 120 may be connected to the network 104, and other components of the virtualized computing system 110 may communicate with other components of the virtualized environment 100 via the front end 120.

The user may use the virtual computing system 110 to execute user code thereon. For example, a user may wish to run a piece of code in conjunction with a web or mobile application that the user has developed. One way to run code would be to obtain a virtual machine instance from a service provider that provides infrastructure as a service, configure the virtual machine instance to meet the user's needs, and run code using the configured virtual machine instance. Alternatively, the user may send a code execution request to the virtual computing system 110. The virtual computing system 110 may handle the acquisition and configuration of computing capacity (e.g., containers, instances, etc., which are described in more detail below) based on the code execution request and execute the code using the computing capacity. The virtual computing system 110 may automatically scale up and down based on volume, thereby relieving the user of having to worry about over-utilizing (e.g., acquiring too little computing resources and suffering from performance problems) or under-utilizing (e.g., acquiring more computing resources than needed to run the code, and thus causing too much payment).

Front end

The front end 120 handles all requests to execute user code on the virtual computing system 110. In one embodiment, the front end 120 acts as the front door for all other services provided by the virtual computing system 110. The front end 120 processes the request and ensures that the request is properly authorized. For example, the front end 120 may determine whether the user associated with the request is authorized to access the user code specified in the request.

User code, as used herein, may refer to any program code (e.g., programs, routines, subroutines, threads, etc.) written in a particular programming language. In this disclosure, the terms "code," "user code," and "program code" may be used interchangeably. Such user code may be executed to perform particular tasks, such as particular web applications or mobile applications developed in connection with a user. For example, the user code may be written in JavaScript (node. js), Java, Python, and/or Ruby. The request may include the user code (or its location) and one or more arguments to be used to execute the user code. For example, a user may provide a user code and a request to execute the user code. In another example, the request may identify previously uploaded program code by name or unique ID (e.g., upload code using API). In yet another example, the code may be included in the request and uploaded to a separate location (e.g., storage service 108 or a storage system internal to virtual computing system 110) before the request is received by virtual computing system 110. The virtual computing system 110 may change its code execution policy based on where code is available when processing the request.

Front end 120 may receive a request to execute such user code in response to a hypertext transfer protocol secure (HTTPS) request from a user. Also, any information (e.g., headers and parameters) included in the HTTPS request may also be processed and utilized when executing user code. As discussed above, any other protocol including, for example, HTTP, MQTT, and CoAP may be used to transmit messages containing code execution requests to the front end 120. When an event is detected, the front end 120 may also receive a request to execute these user codes, such as an event that a user has registered to trigger the generation of an automatic request. For example, the user may have registered a user code with the auxiliary service 106 and specified that whenever a particular event occurs (e.g., uploading a new file), a request to execute the user code is sent to the front end 120. Alternatively, the user may have registered for a timed job (e.g., executing the user code every 24 hours). In such an example, when the scheduled time reaches the timing job, a request to execute the user code may be sent to the front end 120. In yet another example, the front end 120 may have an incoming queue of code execution requests, and the front end 120 may process the user request when the user's batch work is removed from the work queue of the virtual computing system. In yet another example, the request may originate from another component within the virtualized computing system 110 or from another server or service not shown in FIG. 1.

The user request may specify one or more third party libraries (including native libraries) to be used with the user code. In one embodiment, the user request is a ZIP file containing the user code and identification of any libraries (and/or their storage locations). In some embodiments, the user request includes metadata indicating the program code to be executed, the language in which the program code was written, the user associated with the request, and/or the computing resources (e.g., memory, etc.) reserved for executing the program code. For example, the program code may be provided with requests previously uploaded by the user, requests provided by the virtual computing system 110 (e.g., standard routines), and/or requests provided by third parties. In some embodiments, such resource level constraints are specified for a particular user code (e.g., how much memory will be allocated to execute the particular user code) and may not vary with each execution of the user code. In such a case, the virtual computing system 110 may access such resource level constraints prior to receiving each individual request, and the individual requests may not specify such resource level constraints. In some embodiments, the user request may specify other constraints, such as permission data indicating the type of permission the request must execute the user code. The virtual computing system 110 may use such permission data to access private resources (e.g., over a private network).

In some embodiments, the user request may specify actions that should be taken to process the user request. In such embodiments, the user request may include an indicator for enabling one or more execution modes in which user code associated with the user request is to be executed. For example, the request may include a flag or header indicating whether the user code should be executed in a debug mode in which debugging and/or logging output generated in connection with execution of the user code is provided back to the user (e.g., via a console user interface). In such an example, the virtual computing system 110 may examine the request and look for a flag or header, and if present, the virtual computing system 110 may modify the behavior of the container in which the user code is executed (e.g., the logging facility) and cause the output data to be provided back to the user. In some embodiments, the behavior/mode indicator is added to the request through a user interface provided to the user by the virtual computing system 110. Other features, such as source code analysis, remote debugging, etc., may also be enabled or disabled based on the indication provided in the request.

In some embodiments, the virtual computing system 110 may include multiple front ends 120. In such embodiments, a load balancer may be provided to distribute incoming requests to multiple front ends 120, e.g., in a round robin fashion. In some embodiments, the manner in which the load balancer distributes incoming requests to the plurality of front ends 120 may be based on the state of the warm-up pool 130A and/or the active pool 140A. For example, if the capacity in the warm-up pool 130A is deemed sufficient, the request may be distributed to multiple front ends 120 based on the respective capacities of the front ends 120 (e.g., based on one or more load balancing constraints). On the other hand, if the capacity in the warm-up pool 130A is less than the threshold amount, one or more such load balancing restrictions may be removed such that requests may be distributed to the plurality of front ends 120 in a manner that reduces or minimizes the number of virtual machine instances acquired from the warm-up pool 130A. For example, even if requests are routed to front end A according to load balancing constraints, if front end A needs to take an instance from the warm-up pool 130A to service a request, but front end B can use one instance in its active pool to service the same request, the request can be routed to front end B.

Preheating pool manager

When the virtual computing system 110 receives a request to execute user code on the virtual computing system 110, the warm-up pool manager 130 ensures that the virtual machine instance is ready for use by the worker manager 140. In the example shown in FIG. 1, warm-up pool manager 130 manages warm-up pool 130A, which is a group (sometimes referred to as a pool) of pre-initialized and pre-configured virtual machine instances that may be used to service incoming user code execution requests. In some embodiments, the warm-up pool manager 130 causes virtual machine instances to be started on one or more physical computers within the virtual computing system 110 and added to the warm-up pool 130A. In other embodiments, the warm-up pool manager 130 communicates with an auxiliary virtual machine instance service (e.g., instance provisioning service 109 of fig. 1) to create a new instance and add it to the warm-up pool 130A. In some embodiments, the warm-up pool manager 130 may utilize physical computing devices and one or more virtual machine instance services within the virtual computing system 110 to acquire and maintain computing capacity that may be used to service code execution requests received by the front end 120. In some embodiments, the virtual computing system 110 may include one or more logical knobs or switches for controlling (e.g., increasing or decreasing) the available capacity in the pre-heat pool 130A. For example, a system administrator may use such a knob or switch to increase the available capacity (e.g., the number of pre-boot instances) in the pre-heat pool 130A during peak hours. In some embodiments, the virtual machine instances in the warm-up pool 130A may be configured based on a predetermined set of configurations that are not relevant to the particular user request to execute user code. The set of predetermined configurations may correspond to various types of virtual machine instances executing user code. The warm-up pool manager 130 may optimize the type and number of virtual machine instances in the warm-up pool 130A based on one or more metrics related to current or previous user code execution.

As shown in FIG. 1, an instance may have an Operating System (OS) and/or language runtime loaded thereon. For example, a warm-up pool 130A managed by warm-up pool manager 130 includes instances 152, 154. The instance 152 includes an OS 152A and a runtime 152B. Instance 154 includes OS 154A. In some embodiments, the instances in the warm-up pool 130A may also include containers (which may also contain copies of operating systems, runtimes, user code, etc.), which will be described in more detail below. Although instance 152 is shown in FIG. 1 as including a single runtime, in other embodiments, the instance depicted in FIG. 1 also includes two or more runtimes, where each runtime can be used to run different user code. In some embodiments, warm-up pool manager 130 may maintain a list of instances in warm-up pool 130A. The instance list may also specify a configuration of the instance (e.g., OS, runtime, container, etc.).

In some embodiments, the virtual machine instances in the warm-up pool 130A may be used to service any user's requests. In one embodiment, all virtual machine instances in the warm-up pool 130A are configured in the same or substantially similar manner. In another embodiment, the virtual machine instances in the warm-up pool 130A may be configured in different ways to accommodate different user needs. For example, the virtual machine instance may have loaded thereon a different operating system, a different language runtime, and/or a different library. In yet another embodiment, the virtual machine instances in the warm-up pool 130A may be configured in the same or substantially similar manner (e.g., with the same OS, language runtime, and/or library), but some of those instances may have different container configurations. For example, two instances may have runtimes for both Python and Ruby, but one instance may have a container configured to run Python code, while another instance may have a container configured to run Ruby code. In some embodiments, multiple warming pools 130A are provided, each having virtual machine instances configured in the same manner.

The warming pool manager 130 can pre-configure the virtual machine instances in the warming pool 130A such that each virtual machine instance is configured to satisfy at least one of the operating conditions that may be requested or specified by a user request to execute program code on the virtual computing system 110. In one embodiment, the operating conditions may include a programming language in which potential user code may be written. For example, such languages may include Java, JavaScript, Python, Ruby, and the like. In some embodiments, the set of languages that may be used to write the user code may be limited to a predetermined set of languages (e.g., a set of 4 languages, although more or less than four sets of languages are provided in some embodiments) in order to facilitate pre-initialization of virtual machine instances that may satisfy requests to execute the user code. For example, when a user is configuring a request via a user interface provided by the virtual computing system 110, the user interface may prompt the user to specify one of the predetermined operating conditions for executing the user code. In another example, a Service Level Agreement (SLA) for utilizing services provided by the virtual computing system 110 may specify a set of conditions (e.g., programming language, computing resources, etc.) that a user request should satisfy, and the virtual computing system 110 may assume that the request is requested to satisfy the set of conditions when processing the request. In another example, the operating conditions specified in the request may include: an amount of computing capacity to be used to process the request; the type of request (e.g., HTTP versus trigger event); a timeout of the request (e.g., a threshold time after the request may terminate); security policies (e.g., which instances in the pre-heat pool 130A may be controlled for which user; scheduling information (e.g., a time at which the virtual computing system is requested to execute the program code, a time after which the virtual computing system is requested to execute the program code, a time window during which the virtual computing system is requested to execute the program code, etc.), etc.

Worker manager

The worker manager 140 manages instances for servicing incoming code execution requests. In the example shown in FIG. 1, worker manager 140 manages an activity pool 140A, which is a group (sometimes referred to as a pool) of virtual machine instances currently assigned to one or more users. Although the virtual machine instances are described herein as being assigned to a particular user, in some embodiments, an instance may be assigned to a group of users such that the instance is bound to the group of users and any member of the group may utilize resources on the instance. For example, users in the same group may belong to the same security group (e.g., based on their security credentials) such that executing code of one member in a container on a particular instance does not pose a security risk after code of another member has been executed in another container on the same instance. Similarly, the worker manager 140 can assign instances and containers according to one or more policies that dictate which requests can be executed in which containers and which instances can be assigned to which users. An exemplary policy may specify that instances are assigned to a set of users sharing the same account (e.g., an account used to access services provided by the virtual computing system 110). In some embodiments, requests associated with the same group of users may share the same container (e.g., if the user codes associated therewith are the same). In some embodiments, the request does not distinguish between different users of the group and simply indicates the group to which the user associated with the request belongs.

In the example shown in FIG. 1, the user code executes in a separate computing system called a container. A container is a logical unit created within a virtual machine instance using the available resources on the instance. For example, the worker manager 140 can create a new container or locate an existing container in one of the instances in the activity pool 140A based on information specified in the request to execute the user code and assign the container to the request to process execution of the user code associated with the request. In one embodiment, such a container is implemented as a Linux container. The virtual machine instance in the activity pool 140A may have one or more containers created thereon and may have one or more program codes associated with the user loaded thereon (e.g., in one of the containers or in a local cache of the instance).

As shown in FIG. 1, an instance may have an Operating System (OS), language runtimes, and containers. These containers may have loaded thereon separate copies of OS and language runtime and user code. In the example of FIG. 1, the activity pool 140A managed by the worker manager 140 includes instances 156, 157, 158, 159. The example 156 has containers 156A, 156B. The container 156A has loaded thereon the OS 156A-1, the runtime 156A-2, and the code 156A-3. In the depicted example, container 156A has its own OS, runtime, and code loaded therein. In one embodiment, the OS 156A-1 (e.g., its kernel), the runtime 156A-2, and/or the code 156A-3 are shared between the containers 156A, 156B (as well as any other containers not shown in FIG. 1). In another embodiment, the OS 156A-1 (e.g., any code running outside the kernel), the runtime 156A-2, and/or the code 156A-3 are separate copies created for the container 156A and are not shared with other containers on the instance 156. In yet another embodiment, some portions of the OS 156A-1, the runtime 156A-2, and/or the code 156A-3 are shared between containers on the instances 156, and other portions thereof are separate copies unique to the containers 156A. Examples 157 include containers 157A, 157B, 157C. The instance 158 includes containers 158A, 158B and resources 158C. The instance 159 includes a container 159A and a resource 159A-1. The resource 158C may be an instance-specific resource such as memory, disk storage, a database, etc. that is accessible by any container created thereon (or any code executing therein). The resource 159A-1 may be a container-specific resource, such as memory, disk storage, a database, etc., accessible by any code executing therein.

In the example of fig. 1, the size of the container depicted in fig. 1 may be proportional to the actual size of the container. For example, container 156A occupies more space on instance 156 than container 156B does. Similarly, the containers 157A, 157B, 157C may be the same size, and the containers 158A and 158B may be the same size. The dashed boxes labeled "C" shown in instances 158, 159 indicate the space remaining on the instances that can be used to create new instances. In some embodiments, the size of the container may be 64MB or any multiple thereof. In other embodiments, the size of the container may be any arbitrary size that is less than or equal to the size of the instance in which the container was created. In some embodiments, the size of the container may be any arbitrary size that is less than, equal to, or greater than the size of the instance in which the container was created. How much the size of containers may exceed the size of an instance may be determined based on the likelihood that those containers may be utilized beyond the capacity provided by the instance.

Although components inside the containers 156B, 157A, 157B, 157C, 158A, 158B, 159A are not shown in the example of fig. 1, each of these containers may have various operating systems, language runtimes, libraries, and/or user code. In some embodiments, instances may have user code loaded thereon (e.g., in an instance level cache such as resource 158C), and containers within those instances may also have user code loaded therein. In some embodiments, the worker manager 140 may maintain a list of instances in the activity pool 140A. The instance list may also specify a configuration of the instance (e.g., OS, runtime, container, etc.). In some embodiments, the worker manager 140 may access a list of instances (e.g., including the number and type of instances) in the warm-up pool 130A. In other embodiments, worker manager 140 requests computational capacity from warm-up pool manager 130 without knowledge of the virtual machine instances in warm-up pool 130A.

After the front end 120 has successfully processed the request, the worker manager 140 finds the capacity to service the request to execute the user code on the virtual computing system 110. For example, if there is a particular virtual machine instance in the activity pool 140A that has a container with the same user code (e.g., code 156A-3 shown in container 156A) loaded therein, the worker manager 140 may assign the container to the request and cause the user code to execute in the container. Alternatively, if the user code is available in a local cache of one of the virtual machine instances (e.g., stored on instance 158 but not belonging to any separate container), the worker manager 140 may create a new container on such instance, assign the container to the request, and cause the user code to be loaded and executed in the container.

If the worker manager 140 determines that no user code associated with the request is found on any instance in the activity pool 140A (e.g., in a container or local cache of instances), the worker manager 140 may determine that any instance in the activity pool 140A is currently assigned to a user associated with the request and has the computing capacity to process the current request. If such an instance exists, the worker manager 140 may create a new container on the instance and assign the container to the request. Alternatively, the worker manager 140 can also configure existing containers on the instance assigned to the user and assign the containers to the request. For example, the worker manager 140 may determine that if a particular library required by the current user request is loaded on an existing container, the existing container may be used to execute user code. In this case, the worker manager 140 may load specific libraries and user code onto the container and use the container to execute the user code.

If the active pool 140A does not contain any instances currently assigned to the user, the worker manager 140 pulls a new virtual machine instance from the warm-up pool 130A, assigns the instance to the user associated with the request, creates a new container on the instance, assigns the container to the request, and causes user code to be downloaded and executed on the container.

In some embodiments, the virtual computing system 110 is adapted to begin executing user code shortly after the user code is received (e.g., by the front end 120). The time period may be determined as a time difference between initiating execution of the user code (e.g., in a container on a virtual machine instance associated with the user) and receiving a request to execute the user code (e.g., received by the front end). The virtual computing system 110 is adapted to begin executing user code within a time period that is less than a predetermined duration. In one embodiment, the predetermined duration is 500 ms. In another embodiment, the predetermined duration is 300 ms. In another embodiment, the predetermined duration is 100 ms. In another embodiment, the predetermined duration is 50 ms. In another embodiment, the predetermined duration is 10 ms. In another embodiment, the predetermined duration may be any value selected from the range of 10ms to 500 ms. In some embodiments, the virtual computing system 110 is adapted to begin executing user code for a period of time less than a predetermined duration if one or more conditions are met. For example, the one or more conditions may include any one of: (1) upon receiving the request, the user code is loaded onto a container in the activity pool 140A; (2) upon receiving the request, the user code is stored in the code cache of the instance in the activity pool 140A; (3) the activity pool 140A contains instances that are assigned to users associated with the request when the request is received; or (4) the preheat pool 130A has the capacity to handle requests when they are received. In some embodiments, instead of initiating the requested code execution upon receiving a code execution request, the virtual computing system 110 may schedule code execution according to scheduling information provided by the request. For example, the request may specify a time window (e.g., between 3:00AM and 4:00AM on the next monday morning) for which the virtual computing system 110 is requested to execute code execution, and the virtual computing system 110 may schedule code execution based on certain performance considerations (e.g., workload, latency, etc.).

The user code may be downloaded from an auxiliary service 106, such as storage service 108 of fig. 1. The data 108A shown in fig. 1 may include user code uploaded by one or more users, metadata associated with such user code, or any other data used by the virtual computing system 110 to perform one or more of the techniques described herein. Although only storage service 108 is shown in the example of FIG. 1, virtual environment 100 may include other levels of storage systems from which user code may be downloaded. For example, each instance may have one or more storage systems associated with the instance on which the container was created, either physically (e.g., local storage residing on the physical computing system on which the instance is running) or logically (e.g., a storage system in network communication with the instance and providing network attachment within or outside of virtual computing system 110). Alternatively, the code may be downloaded from a web-based data store provided by storage service 108.

Once worker manager 140 finds one of the virtual machine instances in warm-up pool 130A that can be used to service the user code execution request, warm-up pool manager 130 or worker manager 140 takes the instance from warm-up pool 130A and assigns it to the user associated with the request. The allocated virtual machine instance is taken from the warm-up pool 130A and placed in the active pool 140A. In some embodiments, once a virtual machine instance has been assigned to a particular user, the same virtual machine instance cannot be used to service the request of any other user. This provides security benefits to the user by preventing possible user resource mixing. Alternatively, in some embodiments, multiple containers belonging to (or assigned to) different users (or requests associated with different users) may coexist on a single virtual machine instance. This approach may improve the utilization of the available computing capacity. In some embodiments, the virtual computing system 110 may maintain a separate cache in which user code is stored to act as an intermediate level of caching system between the local cache of the virtual machine instance and the network-based network storage (e.g., accessible via the network 104).

After the user code has been executed, the worker manager 140 may tear down the container for executing the user code to free up the resources it occupies for other containers in the instance. Alternatively, the worker manager 140 may keep the container running to use it to service additional requests from the same user. For example, if another request associated with the same user code that has been loaded in the container is received, the request may be assigned to the same container, thereby eliminating the delay associated with creating a new container and loading the user code into the container. In some embodiments, the worker manager 140 may tear down the instance in which the container for executing the user code was created. Alternatively, the worker manager 140 may keep the instance running to use it to service additional requests from the same user. The determination of whether to keep the running of the container and/or instance after the user code completes execution may be based on a threshold time, the type of user, an average amount of requests by the user, periodicity information (e.g., a container/instance in the activity pool 140A on which the user code is not currently executing may be (i) kept active if the periodicity information indicates that additional requests are expected to arrive soon, or (ii) kept terminated if the periodicity information indicates that additional requests may not arrive fast enough to justify keeping the container/instance active), and/or other operating conditions. For example, after a lower threshold time has elapsed without any activity (e.g., execution of code) (e.g., 5 minutes, 30 minutes, 1 hour, 24 hours, 30 days, etc.), the container and/or virtual machine instance is closed (e.g., deleted, terminated, etc.) and the resources allocated to it are released. In some embodiments, the threshold time elapsed before the container is removed is shorter than the threshold time elapsed before the instance is removed.

In some embodiments, the virtual computing system 110 may provide data to one or more auxiliary services 106 as it services incoming code execution requests. For example, the virtual computing system 110 may communicate with a monitoring/logging/billing service 107. The monitoring/logging/billing service 107 may include: a monitoring service for managing monitoring information received from the virtual computing system 110, such as the status of containers and instances on the virtual computing system 110; a logging service for managing log information received from the virtual computing system 110, such as activities performed by containers and instances on the virtual computing system 110; and a billing service for generating billing information associated with executing user code on the virtual computing system 110 (e.g., based on monitoring information and/or log information managed by the monitoring service and the log service). In addition to the system-level activities that may be performed by monitoring/logging/billing service 107 (e.g., on behalf of virtual computing system 110) as described above, monitoring/logging/billing service 107 may provide application-level services on behalf of user code executing on virtual computing system 110. For example, monitoring/logging/billing service 107 may monitor and/or record various input, output, or other data and parameters on behalf of user code executing on virtual computing system 110. Although shown as a single block, the monitoring, logging and billing service 107 may be provided as a separate service.

In some embodiments, the worker manager 140 may perform a health check on the instances and containers managed by the worker manager 140 (e.g., those instances and containers in the activity pool 140A). For example, the health check performed by the worker manager 140 may include determining whether the instances and containers managed by the worker manager 140 have any of the following issues: (1) mis-configured networking and/or boot configurations, (2) exhausted memory, (3) corrupted file systems, (4) incompatible kernels, and/or any other problem that may compromise the performance of instances and containers. In one embodiment, the worker manager 140 performs the health check periodically (e.g., every 5 minutes, every 30 minutes, every hour, every 24 hours, etc.). In some embodiments, the frequency of health checks may be automatically adjusted based on the results of the health checks. In other embodiments, the frequency of health checks may be adjusted based on user requests. In some embodiments, the worker manager 140 may perform similar health checks on instances and/or containers in the pre-heat pool 130A. The instances and/or containers in the pre-heat pool 130A may be managed together with or separately from those in the active pool 140A. In some embodiments, where the health of the instances and/or containers in preheat pool 130A is managed separately from activity pool 140A, preheat pool manager 130, in place of worker manager 140, may perform the above-described health checks on the instances and/or containers in preheat pool 130A.

Route manager

The routing manager 150 creates and manages a mapping between incoming code execution requests received by the virtual compute system 110 and the computing capacity used to process those code execution requests. The mapping may facilitate multiple requests to use and reuse certain container-specific resources. For example, a given container may be associated with a cache that stores results of code execution and may be accessed by any program code executing in the given container. Thus, once the code execution results associated with a request are stored in the cache, subsequent requests of the same type may be more efficiently processed using the code execution results stored in the cache if they are routed to a given container. Although the routing manager 150 is shown as a distinct component within the virtual computing system 110, some or all of the functions of the routing manager 150 may be performed by the front end 120, the pre-heat pool manager 130, and/or the worker manager 140. For example, the routing manager 150 may be implemented entirely within one of the other components of the virtual computing system 110 or in a distributed manner on the other components of the virtual computing system 110. In the example of fig. 1, routing manager 150 includes mapping data 150A. The mapping data 150A may include data indicating how incoming code execution requests should be routed to containers running on the virtualized computing system 110. Mapping data 150A may also include any container/routing policy specified by a user or determined by routing manager 150 for routing incoming requests received by virtual computing system 110. Mapping data 150A may be stored in storage internal to virtual computing system 110 and/or in external storage (e.g., storage service 108) and periodically backed up.

The routing manager 150 may include: a routing parameter determination unit for determining routing parameters associated with a given code execution request, and a container lookup unit for determining an appropriate container to be used for processing the given code execution request based on the determined routing parameters. An exemplary configuration of the routing manager 150 is described in more detail below with reference to fig. 2.

Maintaining resources associated with a container

In some embodiments, the container running on the virtual computing system 110 may be associated with container-specific resources (such as memory, disk storage, databases, etc. accessible by any program code executing in the container). For example, code execution results of program code executing in a container may be cached in a local memory associated with the container. The virtual computing system 110 may utilize the cached results to facilitate other code execution requests of the same type (e.g., requests associated with the same program code and the same set of arguments). In one example, a user may wish to provide a weather service that allows people to type in zip codes and view weather information associated with those zip codes. Each time a zip code is entered, the entered zip code may be used to send a request to the virtual computing system 110 to perform a weather information lookup function (e.g., in a container created on the virtual machine instance assigned to the user). The lookup function may retrieve weather information from a weather database using a zip code and output the retrieved weather information. The retrieved weather information may also be stored in a memory associated with the container such that any future code execution requests may utilize the stored information. For example, if another request is received with the same zip code (e.g., within a threshold time period), the weather information lookup function may simply return the cached results without going to the weather database, thereby implementing some delay gain. Even if the current weather information is cached, if subsequent requests are routed to a different container that does not have access to the cached information, the cached information will not be available for the weather information lookup function and the latency gain cannot be realized. Additionally or alternatively, the virtual computing system 110 can have other resources, such as memory, disk storage, databases, etc., that can be accessed at an instance level (e.g., accessible by any program code executing on any container on or above a particular instance) and/or at an account level (e.g., accessible by any program code executing on an instance associated with a user).

Mapping between requests and specific containers

In some embodiments, the routing manager 150 processes incoming code execution requests and determines to which containers the requests should be routed. For example, when the virtual computing system 110 first receives a request associated with a given program code, there may not be a mapping that routes the request to a particular container running on the virtual computing system 110. Upon receiving such a request, the instance may be moved from the warm-up pool 130A to the active pool 140A and assigned to the user associated with the request. The virtual computing system 110 may then create a container on the instance, load the program code onto the container, and execute the program code in the container. Any information generated or obtained when executing program code may be stored in one or more computing resources associated with the container. The computing resources may include local memory and/or disk storage that is only accessible to the container. The computing resources may also include databases maintained on the virtual computing system 110 or on external services (e.g., the auxiliary service 106). Upon storing information generated or obtained when executing program code in one or more computing resources, routing manager 150 may update mapping data 150A so that future requests that may benefit from utilizing the stored container-specific information will be routed to that particular container.

The routing manager 150 may update and maintain the mapping data 150A using any known hashing scheme or other data structure. For example, the execution parameters included in the request may be hashed and mapped to the ID of the container. For each program code execution on virtual computing system 110, mapping data 150A may include tables that map various combinations of execution parameters to particular containers. The routing manager 150 may hash one or more enforcement parameters included in the request to determine routing parameters, and use the routing parameters to determine the particular container to which the mapping data 150A maps the routing parameters. For example, if a first request requests performance of the Function Foo with the parameters "current temperature" and "90020," the routing manager 150 may route the first request to a first container, and if a second request requests performance of the Function Foo with the parameters "current temperature" and "10012," the routing manager 150 may route the second request to a second container. In another example, the request may specify a name of a database that the request is configured to manage or modify. In such an example, the database name may serve as a routing parameter that maps to a particular container (e.g., a request associated with database a that maps to container a, a request associated with database B that maps to containers B, C and D, a request associated with databases C, D and E that map to container E, etc.).

In other embodiments, only a subset of the parameters are considered in generating the routing parameters. In such an example, each request associated with the Function Foo and the parameter "current temperature" may be routed to the same container, regardless of the other parameter values. In yet another example, the mapping data 150A may be such that all requests associated with the Function Foo will be routed to a particular container. Although an example of routing to a particular container is used in some embodiments of the present disclosure, the techniques discussed herein may be adapted to provide mapping data 150A that maps requests to particular instances, particular front ends, particular geographic locations, particular worker managers, and/or the like, such that computing resources of a relevant level of granularity (e.g., instance-specific resources, account-specific resources, location-specific resources, and/or the like) may be requested to be utilized.

Mapping data 150A may be provided by a user, for example, via one or more UIs, CLIs, APIs, and/or other programming interfaces. For example, when a container is created or terminated, the virtual computing system 110 may provide certain container information to the user so that the user may provide mapping data about the particular container. In some embodiments, the mapping data 150A may specify that a first set of requests (e.g., requests associated with a Function Foo, requests containing specific parameters specified by a user, requests received during a particular hour, etc.) should be routed to one container and a second set of requests should be routed to another container. By providing an indication in the request that the request will be processed using a particular container, and/or by providing a mapping between parameter values (e.g., hash values) and particular containers, the user can control which containers are used to process which requests (and thus which container-specific resources are accessed), and by omitting expensive operations using container-specific resources (e.g., data stored locally within the container), latency reduction associated with the request is achieved. Mapping data 150A may be based on user accounts (e.g., requests associated with different user accounts routed to different containers), program code (e.g., requests associated with different program code routed to different containers), time of day (e.g., requests routed to different containers received at different hours, days, months, years, etc.), geographic regions (e.g., requests associated with different geographic regions routed to different containers), functional arguments (e.g., requests routed to different containers with different parameters or combinations of arguments), other request parameters or metadata, and so forth. In some embodiments, portions of mapping data 150A that will cause such requests to be routed to this particular container are deleted upon container termination (e.g., due to lack of use).

Managing backend resources

In some embodiments, the routing manager 150 may monitor back-end resources utilized in connection with incoming code execution requests. In some cases, program code executing on the virtual compute system 110 may use resources owned by a third party or user (e.g., disk storage, databases, networks, etc.), and such resources may not scale and those resources are managed by the virtual compute system 110 (e.g., handling large bursty traffic), thereby creating an impedance mismatch. In this case, the virtual computing system 110 may allow the user to specify a mapping between requests and containers so that back-end resources are utilized in a more controlled manner. For example, the mapping data 150A may specify that all requests utilizing a first back-end resource should be routed to a first container and that all requests utilizing a second back-end resource should be routed to any container in a set of ten containers. In such an example, when executing the program code in the first container, the user may determine that the program code is the only one that accesses or manages the first back-end resource, and that the first back-end resource is not overly burdened by a large number of simultaneous requests.

User customization of computing capacity

In some embodiments, the virtual computing system 110 may allow a user to customize the instances for processing their requests. For example, a user may specify, e.g., via one or more UIs, CLIs, APIs, and/or other programming interfaces, that a particular version of a language runtime (or other software component) is to be provided on an instance that services a code execution request associated with the user. In response, when a new instance is assigned to a user, the virtual computing system 110 can install the specified version of the language runtime such that any container created thereon can access the installed language runtime. In some embodiments, routing manager 150 may update mapping data 150A such that requests associated with the user are routed to one or more containers running on these customization instances such that subsequent containers created on the customization instances may take advantage of the customizations after the initial customization (and associated latency hits).

Front end/route manager chat routing information

In some embodiments, the virtual computing system 110 may have multiple front ends 120 associated with different sets of the warming pool manager 130 and the worker manager 140. In such embodiments, the routing manager 150 may communicate with each front end 120 to provide routing information for routing incoming code execution requests to a particular container (or instance). In other embodiments, each front end 120 may be associated with a different routing manager 150. In such embodiments, once new routing information (e.g., mapping data 150A) becomes available, the front end 120 and/or the routing manager 150 may chat with each other so that incoming code execution requests may be routed to the appropriate container. For example, once one of the front end 120 or the route manager 150 obtains or determines new mapping information, the mapping information is shared with all of the front ends 120 or the route manager 150 on the virtual computing system 110.

Account, instance, function, and container level resources (11)

In some embodiments, the virtual computing system 110 maintains one or more computing resources at various levels. For example, the virtual computing system 110 may maintain one or more account level resources (e.g., accessible by execution of any code associated with the same account), one or more instance level resources (e.g., accessible by execution of any code on the same instance), one or more function level resources (e.g., accessible by execution of any code associated with the same function or program code), one or more container level resources (e.g., accessible by execution of any code in the same container), and so forth. The computing resources may include memory, disk storage, databases, networks, or any other resource that may be accessed during execution of the code.

External service for controlling container life cycle

In some embodiments, the virtual compute system 110 may allow third party services (e.g., application performance monitoring services, logging services, etc.) some amount of control over the lifecycle of containers created on the virtual compute system 110. In some cases, these services may have the ability to only ingest certain types of data or only ingest certain rates. In such a case, the virtual computing system 110 may allow the services to control the speed at which the container is created and/or terminated, for example, by allowing the services to perform one or more operations before/after the container is created or terminated. In one example, the logging service may perform certain logging operations at the end of the life cycle of the container to perform logging and cleaning operations. In another example, the external service may specify that only certain types of requests should be sent to the container managed by the external service. In response, the routing manager 150 can update the mapping data 150A such that only requests of a specified type are routed to containers managed by the external service.

Container lifecycle management

In some embodiments, the routing manager 150 may keep one or more containers active by avoiding terminating a container immediately or shortly after code execution in the container has completed. For example, the routing manager 150 may cause certain containers to remain active even if the containers are not available to service any existing code execution requests. The decision whether to keep such containers active may be based on the expected frequency with which the virtual computing system 110 receives requests to be processed using the containers and/or the amount of information already stored in the container-specific computing resources in association with the containers. The process of keeping a container alive beyond its normal schedule is described in more detail below with reference to fig. 6.

Information about future code execution

In some embodiments, the routing manager 150 may maintain execution information regarding future code executions on the virtual compute system 110. For example, such information may specify that function Y should be executed upon successful execution of function X, and that function Z should be executed upon failure of execution of function X. In such an example, the routing manager 150 may check the execution information after each attempt to execute the program code to determine the next course of action. In another example, the routing manager 150 may maintain execution information in one or more computing resources associated with a container in which the program code is executed and check the execution information after the request has been routed to the container (e.g., just as the routing manager 150 would check cached information to determine whether or how the program code should be executed).

Local timer

In some embodiments, the routing manager 150 may maintain local timers within a single container. For example, the first container may maintain a local timer in a computing resource (e.g., memory, disk storage, etc.) associated with the first container, and any request routed to the first container may utilize the local timer to execute code in the first container. By maintaining a local timer that may be utilized for requests routed to the first container, the routing manager 150 eliminates the need to access certain external services (e.g., scheduling services external to the virtualized computing system 110) to monitor timing information (e.g., requests may simply utilize the local timer), thereby achieving latency gains.

General architecture for routing manager

FIG. 2 depicts a general architecture of a computing system (referred to as a routing manager 150) that manages virtual machine instances in a virtual computing system 110. The general architecture of the routing manager 150 depicted in fig. 2 includes an arrangement of computer hardware and software modules that may be used to implement aspects of the present disclosure. The routing manager 150 may include more (or fewer) elements than shown in fig. 2. It is not necessary, however, that all of these generally conventional elements be shown in order to provide a workable disclosure. As shown, the routing manager 150 includes a processing unit 190, a network interface 192, a computer-readable medium drive 194, an input/output device interface 196, all of which may communicate with each other over a communication bus. Network interface 192 may provide connectivity to one or more networks or computing systems. Processing unit 190 may thus receive information and instructions from other computing systems or services via network 104. Processing unit 190 may also communicate to and from memory 180, and also provide output information for an optional display (not shown) via input/output device interface 196. The input/output device interface 196 may also accept input from an optional input device (not shown).

Memory 180 may contain computer program instructions (grouped into modules in some embodiments) that processing unit 190 executes to implement one or more aspects of the present disclosure. Memory 180 typically includes RAM, ROM, and/or other permanent, auxiliary, or non-transitory computer-readable media. Memory 180 may store an operating system 184 that provides computer program instructions for use by processing unit 190 in the general management and operation of routing manager 150. Memory 180 may also include computer program instructions and other information for implementing aspects of the present disclosure. For example, in one embodiment, memory 180 includes a user interface unit 182 that generates a user interface (and/or instructions thereof) for display on a computing device, e.g., via a navigation and/or browsing interface, such as a browser or application installed on the computing device. Additionally, memory 180 may include and/or may be in communication with one or more data repositories (not shown), for example, to access user program code and/or libraries.

In addition to and/or in conjunction with the user interface unit 182, the memory 180 may include a routing parameter determination unit 186 and a container lookup unit 188 that may be executed by a processing unit 190. In one embodiment, the user interface unit 182, the routing parameter determination unit 186, and the container lookup unit 188, individually or collectively, implement various aspects of the present disclosure, such as maintaining routing information for routing requests to appropriate containers, processing incoming code execution requests, determining routing parameters associated with requests, identifying containers to be used for processing requests, and so forth, as described further below.

The routing parameter determination unit 186 determines the routing parameters associated with a given code execution request. For example, the routing parameter determination unit 186 processes the code execution request and extracts the routing parameters included in the request. Additionally or alternatively, the routing parameter determination unit 186 determines the routing parameters based on one or more of an account associated with the request, program code associated with the request, execution parameters included in the request, or other metadata associated with the request (such as time of receipt, etc.).

The container lookup unit 188 uses the determined routing parameters to lookup the appropriate container to which the request should be routed. Any known hashing and/or lookup scheme may be used to associate the routing parameters with a particular container and determine the appropriate container based on the routing parameters.

Although the routing parameter determination unit 186 and the container lookup unit 188 are shown in fig. 2 as part of the routing manager 150, in other embodiments all or a portion of the routing parameter determination unit 186 and the container lookup unit 188 may be implemented as a single unit, separate units, or in a distributed manner by other components of the virtual computing system 110 and/or another computing device. For example, in certain embodiments of the present disclosure, another computing device in communication with the virtual computing system 110 may include several modules or components that operate similarly to the modules and components shown as part of the routing manager 150.

Exemplary routines for routing code execution requests

Referring now to FIG. 3, a routine 300 implemented by one or more components of the virtualized computing system 110 (e.g., the routing manager 150) will be described. While the routine 300 is described with respect to implementation by the routing manager 150, one skilled in the relevant art will appreciate that alternative components may implement the routine 300 or that one or more blocks may be implemented by different components or in a distributed manner.

At block 302 of the illustrative routine 300, the routing manager 150 maintains a plurality of virtual machine instances. The plurality of virtual machine instances may include a warm-up pool (e.g., warm-up pool 130A) that includes virtual machine instances on which one or more software components are loaded and waiting to be allocated to a user; and an activity pool (e.g., activity pool 140A) that includes virtual machine instances currently assigned to one or more users. The virtual machine instances in the active pool may have created thereon one or more containers for executing program code therein.

At block 304, the routing manager 150 processes the request to execute the program code on the virtual computing system 110. The request may include a user indication that the program code is to be executed based on one or more computing resources modified by a previous execution of the program code. For example, a previous execution of program code may have caused certain data (e.g., key-value pairs stored in a database, function return values, etc.) to be retrieved (e.g., from an external database service in communication with the virtual computing system 110) or computed and stored locally within the container in which the previous execution occurred, and any subsequent requests that may benefit from accessing the stored data or any subsequent requests that include a user indication to use such stored data may be routed to the container in which the previous execution occurred so that subsequent executions may access the locally stored data. The request may also include user account information (e.g., identification of an account associated with the user), program code information (e.g., identification of the first program code to be executed), and one or more parameters (e.g., execution parameters, function arguments, etc.) to be used to execute the first program code.

At block 306, the routing manager 150 identifies a container associated with the one or more computing resources on a first virtual machine instance of the plurality of virtual machine instances based on a user indication that the program code is to be executed based on the one or more computing resources modified by a previous execution of the program code. For example, the routing manager 150 may calculate the hash value based on the user indication provided in the request and perform a lookup of the hash value in a mapping table (e.g., mapping data 150A) maintained on the virtual computing system that maps the hash value to a particular container. For example, the user indication (or mapping data 150A maintained on the virtual computing system 110) may also specify that requests associated with the same parameter (e.g., a zip code to be used to look up weather information) should be routed to the same parameter container. Although a container is used as an exemplary destination, the request or mapping data 150A may indicate that the request should be routed to one of multiple containers or a group of containers.

At block 308, the routing manager 150 causes the program code to execute in the container. In one example, the routing manager 150 (or another component of the virtual computing system 110, such as the worker manager 140) can determine whether the same type of request has been processed in the container and, if so, whether one or more computer resources associated with the container contain information that can facilitate execution of the program code. For example, results of previous executions associated with the program code may have been stored in a memory associated with the container. In this case, the current execution of the program code may output or return a previously executed result or omit certain computational operations (e.g., data retrieval, function calls, data computation, etc.) based on the availability of the previously executed result, thereby achieving reduced latency.

While the routine 300 of FIG. 3 has been described above with reference to blocks 302-308, the embodiments described herein are not so limited and one or more blocks may be omitted, modified, or exchanged without departing from the spirit of the present disclosure.

Example mapping tables

Referring now to FIG. 4, an exemplary mapping table maintained by the virtualized computing system 110 or the auxiliary service 106 of FIG. 1 will be described. As shown in FIG. 4, table 400 includes incoming code execution requests labeled "request A", "request B", "request C", and "request D", where each request is associated with a routing parameter and a container ID. For example, request a has a routing parameter value of X and a container ID of 2, request B has a routing parameter value of Y and a container ID of 7, request C has a routing parameter value of C and a container ID of 2, and request D has a routing parameter value of D and a container ID of 1. In the example of FIG. 4, assume that requests A-D are received by the virtual computing system 110 sequentially (e.g., A, then B, then C, then D). The routing manager 150 may determine routing parameters associated with request a based on routing parameters included in the request or based on the nature of the request (e.g., associated user account, function, etc.). The routing manager 150 determines that the routing parameter for request a is "X" and identifies the container associated with routing parameter X, which is a container with an ID value of 2. After processing request a, request B is received and routed to a container having an ID value of 7 based on the determined routing parameter value "Y". When request C is received, the routing manager 150 determines that request C has the same routing parameter values as previously processed request a. According to mapping table 400, request C is also processed using a container with an ID value of 2, potentially achieving some delay gain if it is able to utilize the information processed by request a and stored in association with the container with an ID value of 2. Then, based on the determined routing parameter value "Z", request D is received and routed to a container with an ID value of 1. In some embodiments, the routing parameter value may indicate a particular resource requested for execution of the program code (e.g., a resource ID identifying the resource, such as resource 158C and resource 159A-1 of fig. 1). For example, requests requesting use of the same resource may be routed to the same container or group of containers.

The mapping table (or mapping data 150A) is not limited to the configuration shown in the example of fig. 4 and may include any number of parameters that may be used to determine how requests should be routed to which containers.

Exemplary routines for caching code execution results

Referring now to FIG. 5, a routine 500 implemented by one or more components of the virtualized computing system 110 (e.g., the routing manager 150) will be described. While the routine 500 is described with respect to implementation by the routing manager 150, one skilled in the relevant art will appreciate that alternative components may implement the routine 500 or that one or more blocks may be implemented by different components or in a distributed manner.

At block 502 of the illustrative routine 500, the routing manager 150 routes a request to execute program code to a container. The container to which the request is routed may have been identified according to the routine 300 described with reference to FIG. 3.

At block 504, the routing manager 150 determines whether the data is executable with previous code that may be requested to be utilized in one of the resources (e.g., memory, disk storage, database, etc.) associated with the container. In some examples, the routing manager 150 may determine whether the computing resources associated with the container contain any information (e.g., code execution results, or other data, determinations, and/or calculations) that is available for use in connection with the current request. If the routing manager 150 determines that previous code execution data that may be requested to be utilized is available, the routine proceeds to block 510, where the previous code execution data is presented to the program code. Otherwise, the routine 500 proceeds to block 506.

At block 506, the routing manager 150 causes the program code to execute in the container. The program code may execute using one or more parameters included in the request. In some embodiments, the data presented to the cache of the program code causes the program code to start in a different state than the state the program code had when there was no cached data. The program code may omit certain computing operations (such as data retrieval or computation) based on the availability of cached data. Once code execution is complete, the routing manager 150 may store any information that may facilitate processing future requests of the same or similar type in one or more computing resources associated with the container. At block 508, the routing manager 150 outputs code execution results obtained from executing the program code. In some cases, the output code execution results may be the same as the cached data presented to the program code at block 510. In other cases, the program code also processes or modifies the cached data to produce a different code execution result than the cached data.

Although the routine 500 of FIG. 5 has been described above with reference to blocks 502-510, the embodiments described herein are not so limited and one or more blocks may be omitted, modified, or exchanged without departing from the spirit of the present disclosure.

Exemplary routines for managing a Container lifecycle

Referring now to FIG. 6, a routine 600 implemented by one or more components of the virtualized computing system 110 (e.g., the routing manager 150) will be described. While the routine 600 is described with respect to implementation by the routing manager 150, one skilled in the relevant art will appreciate that alternative components may implement the routine 600 or that one or more blocks may be implemented by different components or in a distributed manner.

At block 602 of the illustrative routine 600, the routing manager 150 completes executing the program code on the container. At block 604, the routing manager 150 updates one or more computing resources associated with the container based on the completed execution. The updated computing resources may include information that may facilitate processing future requests of the same or similar type, as discussed with reference to fig. 5.

At block 606, the routing manager 150 determines whether one or more threshold criteria for keeping the container active are met. In some cases, a container may not close immediately even though the virtual computing system 110 does not have any requests that can be processed using the container, in anticipation that such requests may arrive soon and make use of the information stored in association with the container. In this case, if the performance gain realized by keeping the container active and processing future requests using the container exceeds the cost of keeping the container active, the routing manager 150 may select to keep the container active for an extended period of time beyond the normal schedule. The threshold criteria may include a threshold frequency level for receiving requests of the same type. For example, if the container has cached information that may be heavily used in the near future (e.g., expected to be received at a frequency higher than a threshold frequency level), the routing manager 150 may keep the container active. The threshold criteria may include a threshold size of information stored in association with the container. If the container has a large amount of information stored in local memory or storage (e.g., greater than a threshold amount), the routing manager 150 may keep the container active. If the routing manager 150 determines that the threshold criteria are not met, the routine 600 proceeds to 610. Otherwise, the routine 600 proceeds to block 608.

At block 608, the routing manager 150 keeps the container running to handle future code execution requests. In some embodiments, the duration that a container remains active beyond its normal schedule is proportional to the size of the information stored in the container or the expected frequency of receiving requests that will use the information stored in the container. In other embodiments, the container remains active for a fixed duration of time, and the container is closed if the container remains idle for the fixed duration of time.

At block 610, the routing manager 150 causes any cached data associated with the container to be presented to the closing program code (or closing hook) that is configured to execute when the container is closed. The closing program code may be used to perform any logging, monitoring, cleaning, or other operations associated with the container. At block 612, the routing manager 150 causes the shutdown program code to execute in the container. At block 614, the routing manager 150 causes the container to close. In some embodiments, the routing manager 150 may notify the container closure before performing the closure, so that any data stored in the container-specific computing resources may be stored in available persistent storage even after the container is closed.

Although the routine 600 of FIG. 6 has been described above with reference to blocks 602-608, the embodiments described herein are not so limited and one or more blocks may be omitted, modified, or exchanged without departing from the spirit of the present disclosure. For example, in some embodiments, at block 614, the container is not closed, but is reset or cleared.

Other considerations

Those skilled in the art will appreciate that all of the functions described in this disclosure can be implemented in software for execution by the disclosed components and one or more physical processors of a mobile communication device. The software may be persistently stored in any type of non-volatile storage.

Conditional language such as "can," "might," or "may" is generally intended to indicate that certain embodiments include certain features, elements and/or steps, even though other embodiments do not. Thus, such conditional language is not generally intended to imply any way that features, elements and/or steps are required for one or more embodiments or that one or more embodiments necessarily include provisions for deciding, with or without user input or prompting, whether such features, elements and/or steps are included or are to be performed in any particular embodiment.

Any process descriptions, elements, or blocks in flow diagrams described herein and/or depicted in the drawings should be understood as potentially representing modules, segments, or portions of code which include one or more executable instructions for implementing specific logical functions or steps in the process. Alternative implementations are included within the scope of the embodiments described herein in which elements or functions may be deleted, performed out of the order shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art. It should also be appreciated that the above-described data and/or components may be stored on a computer-readable medium and loaded into the memory of a computing device using a drive mechanism associated with a computer-readable storage medium (such as a CD-ROM, DVD-ROM, or network interface) that stores the computer-executable components. Further, the components and/or data may be included in a single device or distributed in any manner. Accordingly, a general purpose computing device may be configured to implement the processes, algorithms, and methods of the present disclosure by processing and/or executing the various data and/or components described above.

The foregoing may be better understood in view of the following clauses:

1. a system for providing low-latency computing capacity, the system comprising:

an external database service configured to store a plurality of key-value pairs;

a virtual computing system comprising one or more hardware computing devices executing particular computer-executable instructions, the virtual computing system in communication with the external database service and configured to at least:

maintaining a plurality of virtual machine instances on one or more physical computing devices, wherein the plurality of virtual machine instances comprises:

a warm-up pool comprising a virtual machine instance having one or more software components loaded thereon and waiting to be allocated to a user; and

an activity pool comprising virtual machine instances currently assigned to one or more users;

processing a first request to execute first program code associated with a first user on the virtual computing system, the first request including an identification of the first user and one or more first execution parameters to be used to execute the first program code;

executing the first program code in a first container created on a first virtual machine instance in the active pool using the one or more first execution parameters, the execution of the program code causing one or more key-value pairs of the plurality of key-value pairs to be retrieved from the external database service and stored locally within the first container;

processing a second request to execute the first program code on the virtual computing system, the second request including the identification of the first user, one or more second execution parameters to be used to execute the first program code, and a user indication to process the second request using the first container previously used to process the first request; and is

Based on the user indication that the second request is to be processed using the first container, executing the first program code in the first container using the one or more second execution parameters, the execution of the first program code accessing the one or more key value pairs stored locally in the first container such that latency associated with executing the first program code based on the one or more second execution parameters is reduced.

2. The system of clause 1, wherein the virtual computing system is configured to identify the first container by computing a hash value associated with the user indication and performing a lookup of the computed hash value in a mapping table maintained by the virtual computing system that maps one or more hash values to a particular container running on the plurality of virtual machine instances.

3. The system of clause 1, wherein executing the first program code using the one or more second execution parameters omits at least one data retrieval from the external database service based on the one or more key value pairs stored locally within the first container.

4. The system of clause 1, wherein the virtual computing system is further configured to keep the first container running even after execution of the first program code using the one or more second execution parameters has been completed, such that a subsequent request can be processed using the first container, the subsequent request providing an indication to use the first container.

5. A system, comprising:

a virtual computing system comprising one or more hardware computing devices executing specific computer-executable instructions and configured to at least:

maintaining a plurality of virtual machine instances on one or more physical computing devices;

processing a first request to execute a first program code on the virtual computing system, the first request including user account information, one or more parameters to be used to execute the first program code, and a user indication that the first program code is to be executed based on one or more computing resources modified by a previous execution of the first program code;

identifying, based on the user indication that the first program code is to be executed based on one or more computing resources modified by a previous execution of the first program code, a first container on a first virtual machine instance of the plurality of virtual machine instances associated with the one or more computing resources modified by the previous execution of the first program code; and is

Causing execution of the first program code in the first container based on the one or more computing resources associated with the first container.

6. The system of clause 5, wherein the virtual computing system is configured to maintain a mapping table that maps one or more values associated with an incoming code execution request to a particular container running on the plurality of virtual machine instances.

7. The system of clause 5, wherein the one or more computing resources associated with the first container comprise local storage within the first container that stores data retrieved from an external service in communication with the virtual computing system in conjunction with the previous execution of the first program code.

8. The system of clause 5, wherein the virtual computing system is further configured to:

determining that the one or more computing resources associated with the first container contain cached data associated with the previous execution of the first program code; and is

Causing presentation of the cached data to the execution of the first program code in the first container, wherein the execution of the first program code omits at least one computing operation based on the cached data, thereby enabling reduced latency.

9. The system of clause 5, wherein the first virtual machine instance has one or more instance-specific computing resources accessible by one or more containers running on the first virtual machine instance.

10. The system of clause 5, wherein the user indication specifies a parameter that causes the virtual computing system to route the first request to a user-selected container.

11. The system of clause 5, wherein the virtual computing system is configured to route incoming code execution requests that include the same user indication to the same container.

12. The system of clause 5, wherein the virtual computing system is further configured to automatically execute the closing program code associated with the first container when the first container is closed.

13. A computer-implemented method, comprising:

as implemented by one or more computing devices configured with specific executable instructions,

identifying, based on the user indication that the first program code is to be executed based on one or more computing resources modified by a previous execution of the first program code, a first container on a first virtual machine instance of the plurality of virtual machine instances associated with the one or more computing resources modified by the previous execution of the first program code; and

14. The method of clause 13, further comprising maintaining a mapping table that maps one or more values associated with the incoming code execution request to a particular container running on the plurality of virtual machine instances.

15. The method of clause 13, wherein the one or more computing resources associated with the first container comprise local storage within the first container that stores data retrieved from an external service in communication with the virtual computing system in conjunction with the previous execution of the first program code.

16. The method of clause 13, further comprising:

determining that the one or more computing resources associated with the first container contain cached data associated with the previous execution of the first program code; and

17. The method of clause 13, wherein the first virtual machine instance has one or more instance-specific computing resources accessible by one or more containers running on the first virtual machine instance.

18. The method of clause 13, wherein the user indication specifies a parameter that causes the virtual computing system to route the first request to a user-selected container.

19. The method of clause 13, further comprising routing an incoming code execution request including the same user indication to the same container.

20. The method of clause 13, further comprising automatically executing closing program code associated with the first container when the first container is closed.

It should be emphasized that many variations and modifications may be made to the above-described embodiments, the elements of which are to be understood as being other acceptable examples. All such modifications and variations are intended to be included herein within the scope of this disclosure and protected by the following claims.

Claims

processing a first request to execute first program code associated with a first user on the virtual computing system, the first request including routing parameters usable to route the first request to computing capacity usable to execute the first program code, and one or more first execution parameters to be used to execute the first program code;

processing a second request to execute the first program code on the virtual computing system, the second request including the same routing parameters previously included in the first request and one or more second execution parameters to be used to execute the first program code; and is

Based on the routing parameters included in the second request, cause the second request to be routed to the first container to execute the first program code in the first container using the one or more second execution parameters, the execution of the first program code accessing the one or more key value pairs stored locally within the first container such that latency associated with executing the first program code based on the one or more second execution parameters is reduced.

2. The system of claim 1, wherein the virtual computing system is configured to identify the first container by computing a hash value of the routing parameter and performing a lookup of the computed hash value in a mapping table maintained by the virtual computing system that maps one or more hash values to a particular container running on the plurality of virtual machine instances.

3. The system of claim 1, wherein executing the first program code using the one or more second execution parameters omits at least one data retrieval from the external database service based on the one or more key value pairs stored locally within the first container.

4. The system of claim 1, wherein the virtual computing system is further configured to keep the first container running even after execution of the first program code using the one or more second execution parameters has completed, such that a subsequent request can be processed using the first container, the subsequent request providing an indication to use the first container.

5. A computer system, comprising:

processing a first request to execute a first program code on the virtual computing system, the first request including a routing parameter usable to route the first request to computing capacity usable to execute the first program code;

determining, based on routing parameters included in the first request, that the first request is routed to a first container that is on a first virtual machine instance of the plurality of virtual machine instances and that is associated with one or more computing resources in the first container that were modified by previous execution of the first program code; and is

Causing the first request to be routed to the first container to execute the first program code in the first container based on the one or more computing resources associated with the first container.

6. The system of claim 5, wherein the virtual computing system is configured to maintain a mapping table that maps one or more values associated with incoming code execution requests to particular containers running on the plurality of virtual machine instances.

7. The system of claim 5, wherein the one or more computing resources associated with the first container comprise local storage within the first container that stores data retrieved from an external service in communication with the virtual computing system in conjunction with the previous execution of the first program code.

8. The system of claim 5, wherein the virtual computing system is further configured to:

9. The system of claim 5, wherein the first virtual machine instance has one or more instance-specific computing resources accessible by one or more containers running on the first virtual machine instance.

10. The system of claim 5, wherein the routing parameters cause the virtual computing system to route the first request to parameters of a user-selected container.

11. The system of claim 5, wherein the virtual computing system is configured to route incoming code execution requests that include the same routing parameters to the same container.

12. The system of claim 5, wherein the virtual computing system is further configured to automatically execute a closing program code associated with the first container when the first container is closed.

13. A computer-implemented method, comprising:

processing a first request to execute a first program code on a virtual computing system, the first request including a routing parameter usable to route the first request to computing capacity usable to execute the first program code;

determining, based on routing parameters included in the first request, that the first request is routed to a first container that is on a first virtual machine instance of the plurality of virtual machine instances and that is associated with one or more computing resources in the first container that were modified by previous execution of the first program code; and

14. The method of claim 13, further comprising maintaining a mapping table that maps one or more values associated with an incoming code execution request to a particular container running on the plurality of virtual machine instances.

15. The method of claim 13, wherein the one or more computing resources associated with the first container include local storage within the first container that stores data retrieved from an external service in communication with the virtual computing system in conjunction with the previous execution of the first program code.